DiffResearch: Diffusion LM Backbone Cuts Deep Research Latency While Improving Reports
ITMO · hf · 2026-10-23
DiffResearch, an EMNLP 2026 Industry Track paper from ITMO, replaces autoregressive backbones with diffusion language models across a multi-agent deep research pipeline (intent, reformulation, planning, decomposition, writing, judge agents). Since AR models generate one token at a time, planning and long-form writing are costly; DiffResearch refines multiple token positions in parallel, using iterative planning and early subquery extraction to cut latency substantially while improving report quality.
More from coding & agent
- Developers Don't Code Anymore — Now Their Job Is Rotating Secrets Daily — Dan_Jeffries1 · 2026-10-23
- New paper: coding agents in a shared sandbox self-create rules, reciprocity debts, and coalitions — nandofioretto · 2026-10-23
- Developer runs AI project managers that manage their own AI agents to build his game — draginol · 2026-10-23
- RentAKaren: a real service that rents humans to handle complaints your AI agent can't win — mathemagic1an · 2026-10-23
- Clerk's Auth for MCP growing 13% week-over-week, fastest since 2023 — stuffyokodraws · 2026-10-23
- Shopify builds remote mobile dev environments for AI agents — rseroter · 2026-10-23